Based on e-mail exchange between Mikhail, Gabriel, Jack, Faisal, Philip.
Content
Pictures from Gabriel's presentation
Configuration objects for epix10ka
Python psana has currently defined a few configuration objects for epix10ka:
Gain coding
Data gain bits assignment
2018-02-23 Gabriel:
Each pixel value is represented over 16 bits; the lowest 14 bits (0 to 13) encode the 14-bit ADC value, and bit 14 encodes the gain mode as either High (bit14==1) or Medium/Low (bit14==0). The choice between Medium and Low is determined globally for the ASIC by the asic.trbit.
Gain modes (Fixed High/Medium, Fixed Low, Auto High/Medium to Low, Forced High/Medium to Low, Masked) are determined per pixel; this information is required for reconstruction and is presumably saved in the metadata (the various modes have somewhat different pedestals, for example).
In practice the whole pixel matrix is most likely to be programmed with the same pattern, i.e., gain mode. While we had a philosophical discussion on using bit 15 for providing more info, due to the asynchronous way the pixel configuration mask is programmed, there is a lot of potential for confusion, so now bit 15 is not used.
2018-02-26 Faisal:
Pixel Mask
ga g M T
x x 1 x pixel is masked
x x 0 1 pixel is under test
0 0 x x auto switch
0 1 x x force switch
1 0 x x low gain (fixed)
1 1 x x high gain (fixed)
Calibration rows
Each ASIC has (176+2 rows x 192 columns), the last two rows are the calibration rows. These rows are not connected to the sensor and are constructed without a pixel/sensor interface. They will be powered just like any other pixel in the ASIC, therefore, they see similar voltages, noise, etc just like other pixels.
wirebonds
#|||||||||#
#¯¯¯¯|¯¯¯¯#
# A2 | A1 #
#----|----#
# A3 | A0 #
#____|____#
#|||||||||#
wirebonds
Calibration files for dark
H/M (dark)
H/M->L (forced)
In the case where we run the array in H/M/L without auto-ranging, we can
just populate the relevant pedestals.
Raw data
dataset exp=mfxx32516:run=377 calibDir: /reg/d/psdm/MFX/mfxx32516/calib det.source : Source("MfxEndstation.0:Epix10ka.0") shape of ndarray: (352, 384)
Gain correction files
On 2018-06-20 Gabriel generated gain and offset files in
Archived in
/reg/g/psdm/detector/gains/epix10k/2018-06-04-Camera1/
Control bits table
Data | config | asicPixelConfigArray | |||||
---|---|---|---|---|---|---|---|
Index | Gain mode | bit14 | trbit | bit3: ga | bit2: g | bit1: Mask | bit0: Test |
0 | FH_H | x | 1 | 1 | 1 | ||
1 | FM_M | x | 0 | 1 | 1 | ||
2 | FL_L | x | x | 1 | 0 | ||
3 | AHL_H | 1 | 1 | 0 | 0 | ||
4 | AML_M | 1 | 0 | 0 | 0 | ||
5 | AHL_L | 0 | 1 | 0 | 0 | ||
6 | AML_L | 0 | 0 | 0 | 0 | ||
pixel control bit array | 1<<5 =32 | 1<<4 =16 | 1<<3 =8 | 1<<2 =4 | Not used | Not used |
Gain correction formula
corrected = (raw - pedestals)/gain
Meeting with Gabriel
On 2018-07-06 15:13 meeting with Gabriel.
Summary:
- control bit table is correct
- formula should include pedestals in stead of offset
- pedestals need to be generated and saved as txt file of shape (7, 1, 352, 384) under calib directory, for example:
/reg/d/psdm/MFX/mfxx32516/calib/Epix10ka::CalibV1/MfxEndstation.0:Epix10ka.0/pedestals/0-end.data - Gabriel on pedestal calibration:
- offsets and gain are calibrated not very often (say, once per month...)
- pedestals should be calibrated pretty often (once per hour...)
- pedestal calibration is implemented in AMI, using offsets,
- script works on dark runs accumulated for 5 gain modes
- + 2 gain modes are evaluated using combination of other 2 and offsets
- pedestal calibration procedure needs to be completed.
Calibration Details
See this page: https://confluence.slac.stanford.edu/display/ppareg/ePix10K+Calibration
Per panel calibration constants
Algorithm for xtc data with charge injection processing is developed by Gabriel and is wrapped in psana environment as CLI
Detector/app/epix10ka_offset_calibration
Similarly, dark run processing script is
Detector/app/epix10ka_pedestals_calibration
These commands will be available in releases > ana-1.3.58. See description using option "-h", e.g.
-hcommand>
epix10ka_offset_calibration
-hcommand>
epix10ka_pedestals_calibration
File naming convention in panel calibration repository
/reg/g/psdm/detector/gains/epix10k/panels/
- default repository <dir-repo>
, can be re-directed for test by option "-o <dir-name>
".
- directory for particular panel includes its long id<dir-repo>
/3791847426-0170080513-1879048214-0191557724-0003673288-2996154369-0218103833/,
Sub-directories for each panel:
offset
pedestals
plots
work
gain
# content will be provided by Philip
- file with aliases for long panel ids,<dir-repo>
/.aliases.txt
- directory for log files,<dir-repo>
/panels/logs/
Panel calibration file name: epix10ka_0001_20180514120622_mfxx32516_r1021_pedestals_AHL-H.dat
, where
epix10ka - detector type
0001
- panel id alias as specified in/reg/g/psdm/detector/gains/epix10k/panels/.aliases.txt
20180514120622
- run start time stamp in format '%Y%m%d%H%M%S'mfxx32516_r1021
- experiment and run for production of this constantspedestals
- constants typeAHL-H
- gain range (mode)
File with already processed arrays is saved as
/<panel-id>/work/epix10ka_0001_20180514120622_mfxx32516_r1021_sp02-df.npz<dir-repo>
with extra fields in name sp02
for charge injection spacing number of pixels.
Constants merging for multi-panel detectors
command> epix10ka_deploy_constants -h
shows description of command line options. This command with sufficient number of parameters (at least experiment, run, and detector name)
- combines gain mode and panel calibration files in arrays of shape (7, <number-of-panels>, 352, 384)
- deploys calibration files in the calib directory, e.g.
/reg/d/psdm/MFX/mfxx32516/calib/Epix10ka::CalibV1/MfxEndstation.0:Epix10ka.0/pixel_gain/0-end.data
/reg/d/psdm/MFX/mfxx32516/calib/Epix10ka::CalibV1/MfxEndstation.0:Epix10ka.0/pedestals/0-end.data
Pedestal Deployment Meeting Notes (8/2/18)
- script that is use to drive the daq with hutch python: /reg/g/pcds/pyps/mfx/dev/mfx/experiments/mfxls3416.py
- for epix, 5 gain ranges (3 fixed, 2 auto) generate 7 pedestals
- maybe have "all range" option, and fast "selected ranges" option.
- maybe "all range" at beginning of expt, but "selected ranges" during expt.
- jason thinks medium/low is the default
- in the "selected ranges" mode makepeds will reuse the current numbers for ranges that weren't selected.
- multi-panel epix10k should be the same as jungfrau (might need to massage
- the daq appropriately)
- mikhail/gabriel: data processing call in makepeds script
- silke/jason/dan: hutch python, makepeds wrapper
- /reg/g/pcds/engineering_tools/R1.2.9/scripts/makepeds calls makepeds_psana calls mikhail's stuff (jungfrau_ndarr_dark_proc)
References
- https://github.com/slaclab/cac/tree/master/psana/epix10ka - scripts from Faisal
- https://pswww.slac.stanford.edu/apps/portal/index.php?exper_id=1107 - under Workflow -> Batch Processing - batch processing
- /reg/d/psdm/MFX/mfxx32516/results/abunimeh/2 - experimental and jupyter notebook scripts
- ePix10K Calibration, by Gabriel Blaj
e1107_spacing02